Glossary
Zero-Knowledge Encryption
A design where the service holding your data cannot open it, because the only keys exist on your side. Strong privacy against the provider, paid for with everything the provider can no longer do for you.
General definition
Zero-Knowledge Encryption describes a product architecture, not a specific algorithm: data is encrypted client-side under keys derived from a user passphrase or generated on the device, and only ciphertext reaches the provider’s servers. The provider has zero knowledge of the content; it cannot read it, hand it to a third party, or be breached for it in readable form. Password managers, encrypted cloud storage and encrypted messengers use the term, often as a marketing label for end-to-end encryption applied to stored data. It should not be confused with a zero-knowledge proof, a cryptographic protocol in which one party proves a statement is true without revealing why; the two share a name and nothing else.
The security claim is only as good as the client. Since the provider ships the application that does the encrypting, a malicious or compromised update can exfiltrate keys, which is why open-source clients, reproducible builds and independent audits matter more in a zero-knowledge product than in a conventional one. Metadata is usually outside the envelope: who talks to whom, when, how much and from which device remains visible to the server, because it needs that information to route and store anything at all.
The trade-offs are structural, not implementation bugs:
- No account recovery: a forgotten passphrase or a lost device with no backup means the data is gone; the provider cannot reset it
- No server-side search or indexing; search happens on the device over data it has already decrypted
- No content moderation, spam filtering, data-loss prevention or AI processing on the server, because the server sees only ciphertext
- No supervisory archive: regulators that require a firm to retain and review communications are not satisfied by records the firm itself cannot read
- Multi-device and sharing require key distribution between devices and users, which is where most real-world weaknesses appear
For a consumer who distrusts the provider, those costs are worth paying. For a hospital, a bank or an insurer, the duties to supervise, retain, produce and erase records usually point the other way, toward strong encryption in transit and at rest with keys the organisation controls and a readable message archive.
In the Ethora ecosystem
Ethora’s default is the opposite of zero-knowledge, on purpose: a compliance archive in which the server encrypts data in transit and at rest but can read messages, so retention policies, server-side search, Trust and Safety moderation, the audit trail with immutable log export and per-user erasure all work. On a dedicated or self-hosted deployment the keys and the archive sit in your own cloud account, so the party with knowledge of the content is you, not Ethora.
Since release 2610 an app owner can switch on optional end-to-end encrypted rooms, which is where zero-knowledge properties apply. In such a room OMEMO 2 keys stay on the users’ devices, attachments are sealed client-side with a per-file key, and the server stores ciphertext it cannot open, so it has zero knowledge of message bodies. It still sees sender, room and reply references, because push and routing need them, and the room gives up server-side moderation, search, export and recovery. The feature is off by default and enabled per app precisely because of that list; it also provides forward secrecy for message bodies through the Double Ratchet.